[SOLVED] Internal Oscillator vs External Crystal

[haxan7]

Not really a ccs question, i apologize in advance.
Pic micro-controller (18LF46k22) is to be used in harsh climate of canada.
Used outdoors,lots of serial communication involved, temp ranges from -30 C to +30 C.

Is internal oscillator suitable?
Can 64 Mhz external crystal be used?

[Ttelmah]

No, you can't use a 64MHz external crystal.
You can use an external 64MHz _oscillator_, or a 16Mhz crystal, and the PLL. Look at table 27-6 in the data sheet. The highest speed the HS oscillator (the internal oscillator to use a crystal), is rated to run, is 20MHz.

Now whether you can use the internal oscillator, is something only you can really answer. The point is that 'serial' covers a lot of ground. You could (for instance), go to a synchronous 'serial' link, which then means that the data clock rate is controlled by the master device, so then the frequency of the clock in the slave wouldn't matter. However assuming you are talking about async serial as used over RS232/485 links, then probably not. Generally the total timing error aimed for, should be kept below perhaps 3% maximum. Now there will be some timing error in the baud rate selection itself (depending what this is), so (for instance), you want 115200bps, then the BRG from a 64Mhz clock will give a -0.79% error. - generally high baud rates will have the biggest errors. Now the internal oscillator is specified to give +/-2% accuracy over 0 to 70C, so would 'just' stay inside this area. However if your temperature range goes out to -40C to +125C, with the extended version chips, then the basic accuracy degrades to +/-5%.

It is possible to use the internal oscillator over the extended temperature range by correcting for this. The drift is quite predictable with temperature, and you can tweak the oscillator using the OSCTUNE register. So what you do is have the chip gently adjust the internal oscillator, to correct for the temperature. It means extra code, with you reading the temperature every few minutes, and then just making the adjustment, but it is possible to get the internal oscillator to give better than +/-1% this way.

You can also do this using the serial itself.
I've in the past had systems send 0xAA repeatedly for a few bytes every few minutes, after a long 'break'. The code in the slave device looks for the break, and then times the interval between the edges, and tweaks it's clock then to correct for the error in this timing. Has the advantage that the chip corrects both for it's timing errors, and the ones in the other device as well...

So if by 'serial communication', you mean async serial, then you have the choices of either working out a way of adjusting the clock, or using a more stable clock source.

Best Wishes

[temtronic]

hmm..."harsh climate of canada", eh ? LOL, Yeah, I'm Canadian.

Ok, if you haven't already done so, buy a small freezer and build 'an environmental test chamber'.You might need some extra insulation outside to get down to -30*C though.
I did that back in '85 to test my remote energy control systems BEFORE doing real field tests. Paid for itself the first week! Made it real easy to test from the comfort of my basement, in July, easy to test all sorts of combinations of hardware and software.

ps.I've always used real crystals,far more stabil and reliable than RC circuits.
hth
jay

[haxan7]

Thanks for the detailed reply, i really appreciate it.

[haxan7]

[haxan7]

Common help a brother out

[Ttelmah]

With what?.

[haxan7]

[Ttelmah]

Seriously, one second searching would find the answer to that. It is basic electronics, which is not what this forum is meant to be about....

Downsides slightly slower start up.
Plus sides, less RFI. Less power needed for the oscillator.
PLL's are used in everything today. Your PC clock will be PLL based, Your cellphone, TV etc..
The highest available fundamental crystal is somewhere just over 60Mhz. Think about chips like the DsPIC's running at 120MHz+. Not even possible without the PLL.

[haxan7]

[gpsmikey]